Abstract
In 3D binocular scenes, parallax results in different occlusion relations between near and far surfaces in the two eyes. Monocular (unmatched) regions at the side of binocular surfaces can support a compelling perception of depth, known as "da Vinci" stereopsis. When the monocular region is nasal in the visual field it is seen behind the binocular surface, consistent with occlusion of a partner image in the other eye, and when temporal it can (more controversially) appear in front, consistent with camouflage of a partner image against the binocular surface. In these experiments we examine the latter situation that would seem to require ecologically not only the correct geometry but also a color match of the monocular region and the binocular surface. Naive observers matched the perceived depth of a rectangular monocular tab attached to a circular binocular shape by adjusting the binocular disparity of a small probe. Observers perceived quantitative depth in both occlusion (far) and camouflage (near) arrangements for a monocular black tab attached to a black binocular figure. However, when the tab was red and violated the viewing constraints for camouflage, depth was only perceived for occlusion. In a binocular control experiment, observers matched the depth of a binocular tab that was colored black in one eye and red in the fellow eye, demonstrating that the monocular tab results were not a consequence of binocular rivalry. No depth from camouflage was perceived when the monocular region was a bar next to the binocular region, indicating that attachment to the binocular figure is critical. This reflects contradictions in the literature between for example von Szily (1921) who used tabs and Nakayama and Shimojo (1990) who used bars. We conclude that da Vinci stereopsis not only involves the sophisticated application of occlusion geometry but also novel matching constraints.
Meeting abstract presented at VSS 2012